Long-term individual-based records of mountain goat mortality and terrain use in relation to avalanches in coastal Alaska during 2005-2022

Description

Notes

Methods

Study system. Mountain goats were studied in four separate areas across a broad geographic range in coastal Alaska (5537 km²; Fig. 1 and Supplementary Table 1) from 2005-2022. This area is within the Coast Mountains biogeographic region (56). Mean monthly temperatures range from -2 to 14°C and mean annual precipitation is 1400 mm in Juneau (57), the area's most populous city. Across the region, annual precipitation ranges from 1 to >8 m and winter snowfall ranges from 0.5 to > 3 m of snow water equivalent (58). During the study period, annual snowfall at sea level in Juneau averaged 233 cm with a range of 89-501 cm.

The region is part of the world's largest contiguous coastal temperate rainforest and is composed primarily of Sitka spruce-western hemlock (Picea sitchensis-Tsuga heterophylla) forests at lower elevations (below 450-750 m). At higher elevations, subalpine and alpine habitats dominated by krummholtz forest, low-growing herbaceous meadows, and ericaceous heathlands are widespread and persist to elevations of about 1400 m. The geologic terrain is complex and strongly influenced by terrain accretion and uplift processes (59). The resulting landscape is highly fractured and dominated by steep, rugged topography that is fragmented by active glaciers, icefields, high-volume river systems, and marine waters (59). The avalanche paths in this study extend from sea level to 2000 m and include a variety of aspects as a result of the complex topography of the Coast Mountains.

Mountain goats in this region are widespread and occur at low to moderate densities, typical of northern coastal areas inhabited by the species (55, 60). Populations exhibit a high degree of local-scale population genetic differentiation, with limited movement among geographically discrete mountain complexes (29, 61, 62). Mountain goats are habitat specialists and select steep, rugged terrain near cliffs and exhibit seasonal variation in altitudinal distribution (33, 62, 63). Mountain goats are partially migratory, with some individuals, depending on the study area, residing in alpine and subalpine habitats throughout the year (64, 65). However, most individuals conduct short-distance (5-10 km), seasonal migrations involving annual movements between high-elevation alpine summer habitats and forested, low-elevation wintering areas (63–65). Downslope migrations tend to correspond with the first major snowfall events at high elevation (i.e., mid-October), while upslope migrations are timed with the onset of the spring snow ablation and pre-parturition period (i.e., early May) (63). Individuals in Lynn Canal are highly migratory and, like mountain goats on the Cleveland Peninsula, primarily use low-elevation forested habitat during winter months, while individuals in Klukwan and Baranof more frequently employ mixed-migration strategies, more often utilizing higher-elevation subalpine and alpine habitats where avalanche exposure is greater (63, 64, 66)(Supplementary Fig. 5). Impacts of human development and activity in the study area are, generally, minimal. Nonetheless, low-intensity or localized activities do occur and include regulated hunting, ground- and air-based recreational tourism, timber harvest, and mining (29, 33, 64). The large mammal predator-prey communities in this area are intact and, in addition to mountain goats, key species include moose (Alces alces), Sitka black-tailed deer (Odocoileus hemionus sitkensis), wolves (Canis lupus), coyotes (Canis latrans), black bears (Ursus americanus), brown bears (Ursus arctos) and wolverines (Gulo gulo); though local variation occurs relative to species distribution and abundance (63, 67). 

Mountain goat monitoring. Adult male and female mountain goats were captured using standard helicopter darting techniques (68). During handling all animals were fitted with mortality-sensing very high frequency (VHF) and/or global positioning system (GPS) radio-collars (Telonics Inc., Mesa, AZ). GPS radio-collars were programmed to acquire a GPS location at 6-hour intervals; ancillary activity sensor and temperature measurements were collected over a 15-minute evaluation period commencing at the initiation of the GPS location acquisition attempt. The age of animals was determined by counting horn annuli (69, 70) and, in some cases, cross-validated by examination of tooth eruption patterns (for young animals) (70) and/or cementum analysis of incisors (for deceased animals; Matson's Laboratory, Milltown, MT). Capture and handling procedures complied with all relevant ethical regulations for animal use and were approved by the Alaska Department of Fish and Game Institutional Animal Care and Use Committee (protocols 05‐11, 2016‐25, 0078‐2018‐68, 0039‐2017‐39) and followed the American Society of Mammalogists guidelines (71).

Following capture, animals were typically monitored at least once per month (often multiple times per month) via aerial telemetry to determine whether animals were alive or dead. Survival status was also determined via examination of GPS radio-collar location, activity, and temperature sensor data, an approach that often enabled temporal determination of death within a 6-hour time window. In cases where animals were determined to have died, an initial fixed-wing aerial reconnaissance of the site was conducted and followed up with a ground-based examination to determine the context and causes of death, to the extent possible. Due to safety and logistic considerations, ground-based examinations were typically conducted after initial aerial reconnaissance and determination of death. Due to the delay, it was not always possible to definitively distinguish between non-avalanche related causes of death (i.e. due to scavenging of carcasses). However, avalanche-caused mortality determinations were definitive and associated with carcasses being buried under, or associated with, avalanche debris and located within active avalanche paths. 

Avalanche simulations and mapping. Avalanche hazard indication maps were developed from terrain analysis, downscaled climate model reanalysis, and numerical simulations of avalanche runout dynamics. Object-based image and terrain analyses were used with a digital terrain model (DTM; 5-m resolution) to determine avalanche potential release areas outside of closed canopy, conifer forest areas (23, 72). Dynamically downscaled climate reanalysis (4-km resolution)(73) was used to calculate the maximum snow depth increase over three days in the 1981-2010 climatology, which was used to determine the avalanche release depth for each potential release area. We recognize that biologically meaningful avalanche activity can occur within closed-canopy forests but maintain that such events are very uncommon in southeast Alaska relative to avalanche activity in alpine areas. As such, for this large-scale approach, we assumed that closed canopy, conifer forest areas were not prone to significant avalanche activity and restricted our automated mapping of potential release areas to landcover types outside this designation.  Potential release areas and release depths were then used in the numerical dynamic avalanche model Rapid Mass Movement Simulations (RAMMS)(24) to simulate millions of individual avalanches within the study areas and map avalanche hazard following the large-scale hazard indication modelling approach developed by Bühler et al. (2022)(74). Mapped avalanche hazard zones were further used to confirm that all mortalities classified as avalanche-related were located in avalanche hazard zones.

Mountain goat spatial analyses. Mountain goat GPS radio-collar location data were compiled and subsequently filtered, using methods described by D'Eon et al. (2002)(75) and D'Eon and Delparte (2005)(76), to ensure geolocational accuracy. Using a geographical information system (GIS), mountain goat GPS location data were intersected with avalanche hazard indication maps to determine the relative proportion of time each mountain goat spent in avalanche terrain during months when avalanche mortalities occurred (Oct-May). We defined avalanche terrain as avalanche potential release areas and runout paths combined, as both features comprised equivalent risk to mountain goats. The proportional use of avalanche terrain was calculated for each individual and coded based on whether the individual did or did not die in an avalanche. Monthly and seasonal differences in proportional use avalanche terrain were analyzed in relation to fate using paired student's t-tests, with P < 0.05 denoting statistical significance. 

Mountain goat mortality and survival estimation. As described above, causes of mortality were ascertained for every deceased individual. All causes of mortality were summarized as either being caused by an avalanche or other, non-avalanche related cause(s), including unknown (Supplementary Table 3). Cause-specific mortality was summarized for each population across all years of study as well as by month and study area. Survival of radio-collared animals was calculated for the annual cycle (June-May), at monthly time steps, using the Kaplan-Meier estimator (77). This method allows for staggered entry and exit of newly captured or deceased animals, respectively. While post-capture effects were not evident in our study, we implemented a conservative approach and excluded mountain goats for survival analysis for three days after capture (following Wagler et al. 2022)(78). Survival was estimated using only avalanche-caused mortality cases in order to determine the proportion of radio-marked animals that died due to avalanches (i.e. population-level mortality) for each year and study area. To ensure our sample was representative of the overall adult population, we conducted annual capture events to compensate for mortality losses, and maintain balanced sex and age classes in our sample of marked individuals (79). On average, 11% of study populations were marked and monitored each year (based on mark-resight aerial survey sightability estimation)(60); a large proportion and overall sample size (n = 421 individuals) for deriving reliable estimates of avalanche-related survival (80).

References

56.       A. L. Gallant, J. M. Binnian, J. M. Omernik, M. B. Shasby, Ecoregions of Alaska (U.S. Government Printing Office, Washington D. C., 1995)

57.       J. B. Fellman, S. Nagorski, S. Pyare, A. W. Vermilyea, D. Scott, E. Hood, Stream temperature response to variable glacier coverage in coastal watersheds of Southeast Alaska. Hydrol. Process. 28, 2062–2073 (2014). doi: 10.1002/hyp.9742

58.       C. S. Shanley, S. Pyare, M. I. Goldstein, P. B. Alaback, D. M. Albert, C. M. Beier, T. J. Brinkman, R. T. Edwards, E. Hood, A. MacKinnon, M. V. McPhee, T. M. Patterson, L. H. Suring, D. A. Tallmon, M. S. Wipfli, Climate change implications in the northern coastal temperate rainforest of North America. Climatic Change. 130, 155–170 (2015). doi: 10.1007/s10584-015-1355-9

59.       H. H. Stowell, Geology of Southeast Alaska: Rock and Ice in Motion. (University of Alaska Press, 2006).

60.       K. S. White, G. W. Pendleton, J. N. Waite, "Development of an aerial survey population estimation technique for mountain goats in Alaska." (Final Wildlife Research Report ADF&G/DWC/WRR-2016-9, Alaska Department of Fish and Game, Division of Wildlife Conservation, Juneau, AK, 2016).

61.       A. B. A. Shafer, S. D. Côté, D. W. Coltman, Hot spots of genetic diversity descended from multiple Pleistocene refugia in an alpine ungulate. Evolution. 65, 125–138 (2011). doi: 10.1111/j.1558-5646.2010.01109.x

62.       A. B. A. Shafer, J. M. Northrup, K. S. White, M. S. Boyce, S. D. Côté, D. W. Coltman, Habitat selection predicts genetic relatedness in an alpine ungulate. Ecology. 93, 1317–1329 (2012). doi: 10.1890/11-0815.1

63.       K. S. White, D. P. Gregovich, G. W. Pendleton, N. L. Barten, A. P. Crupi, R. Scott, D. N. Larsen, Modeling resource selection of mountain goats in southeastern Alaska: applications for population management and highway development planning. Proceedings of the Northern Wild Sheep and Goat Council. 18, 32–42 (2012).

64.       K. S. White, D. P. Gregovich, "Mountain goat resource selection in the Haines-Skagway area: Implications for helicopter skiing management." (Wildlife Research Report ADF&G/DWC/WRR–2018–2, Alaska Department of Fish and Game, Division of Wildlife Conservation, Juneau, AK., 2018).

65.       Y. N. Shakeri, K. S. White, J. N. Waite, Staying close to home: Ecological constraints on space use and range fidelity in a mountain ungulate. Ecol. Evol. 11, 11051–11064 (2021). doi: 10.1002/ece3.7893

66.       C. A. Smith, K. J. Raedeke, Group size and movements of a dispersed, low density goat population with comments on inbreeding and human impact. Proceedings of the Northern Wild Sheep and Goat Council. 3, 54–67 (1982).

67.       S. O. MacDonald, J. A. Cook, The land mammal fauna of Southeast Alaska. Can. Field. Nat. 4, 571–598 (1996).

68.       K. S. White, D. E. Watts, K. B. Beckmen, Helicopter‐based chemical immobilization of mountain goats in coastal Alaska. Wildlife Soc. B. 45, 670–681 (2021). doi: 10.1002/wsb.1229

69.       B. L. Smith, Criteria for determining age and sex of American mountain goats in the field. J. Mammal. 69, 395–402 (1988). doi: 10.2307/1381400

70.       S. M. Brandborg, Life history and management of the mountain goat in idaho. Idaho Wildlife Bulletin, 1–142 (1955).

71.       R. S. Sikes, the Animal Care and Use Committee of the American Society of Mammalogists, 2016 Guidelines of the American Society of Mammalogists for the use of wild mammals in research and education: J. Mammal. 97, 663–688 (2016). doi: 10.1093/jmammal/gyw078

72.       Y. Bühler, D. von Rickenbach, M. Christen, S. Margreth, L. Stoffel, A. Stoffel, R. Kühne, Linking modelled potential release areas with avalanche dynamic simulations: An automated approach for efficient avalanche hazard indication mapping. Proceedings of the International Snow Science Workshop, 810–814 (2018).

73.       R. Lader, A. Bidlack, J. E. Walsh, U. S. Bhatt, P. A. Bieniek, Dynamical downscaling for Southeast Alaska: historical climate and future projections. J. Appl. Meterol. Clim. 59, 1607–1623 (2020). doi: 10.1175/JAMC-D-20-0076.1

74.       Y. Bühler, P. Bebi, M. Christen, S. Margreth, L. Stoffel, A. Stoffel, C. Marty, G. Schmucki, A. Caviezel, R. Kühne, S. Wohlwend, P. Bartelt, Automated avalanche hazard indication mapping on a statewide scale. Nat. Hazards Earth Syst. Sci. 22, 1825–1843 (2022). doi: 10.5194/nhess-22-1825-2022

75.       R. G. D'Eon, R. Serrouya, G. Smith, C. Kochanny, GPS radiotelemetry error and bias in mountainous terrain. Wildlife Soc. B. 30, 430–439 (2002).

76.       R. G. D'Eon, D. Delparte, Effects of radio-collar position and orientation on GPS radio-collar performance, and the implications of PDOP in data screening: Effects of radio-collar position. J. Appl. Ecol. 42, 383–388 (2005). doi: 10.1111/j.1365-2664.2005.01010.x

77.       K. H. Pollock, S. R. Winterstein, C. M. Bunck, P. D. Curtis, Survival analysis in telemetry studies: the staggered entry design. J. Wildl. Manag. 53, 7–15 (1989). doi: 10.2307/3801296

78.       B. L. Wagler, R. A. Smiley, A. B. Courtemanch, G. Anderson, D. Lutz, D. McWhirter, D. Brimeyer, P. Hnilicka, C. P. Massing, D. W. German, T. R. Stephenson, K. L. Monteith, Effects of helicopter net‐gunning on survival of bighorn sheep. J. Wildl. Manag. 86 (2022), doi:10.1002/jwmg.22181.

79.       A. K. Prichard, K. Joly, J. Dau, Quantifying telemetry collar bias when age is unknown: A simulation study with a long-lived ungulate. J. Wildl. Manag. 76, 1441–1449 (2012). doi:10.1002/jwmg.394

80.       D. L. Murray, On improving telemetry-based survival estimation. J. Wildl. Manag. 70, 1530–1543 (2006). doi: 10.2193/0022-541X(2006)70[1530:OITSE]2.0.CO;2